第5章アルゴリズムのハードウェアによる実現：コンピューターシステム設計

Whether you want to be a software developer, a programmer, or a hardware designer, or even to efficiently use any computer system, you need a minimum knowledge of computer organization. You need to know what the different pieces in the computer systems are, how they interact, and how fast each piece is. Knowing the internal workings of computers helps you write more efficient software, design computers of different sizes (like the one in your watch, the one in your phone, or big supercomputers), and make the best decisions when buying a computer.

The programs we use in our computer systems, called applications, have been designed and written by software developers using computer languages, also called high-level languages (HLLs). HLLs have been designed to make the interaction between software developers and computers easier. However, computers do not understand HLLs. Computers understand only 1s and 0s. So, there is a set of programs designed to translate the programs written in an HLL into the 1s and 0s that computers can relate to. The programs written by software developers are translated into a set of instructions such as, “Add number 1 to number 2 and save the result as number 3.” These instructions are stored as a series of 1s and 0s. Each group of those 1s and 0s represents a single instruction. All the instructions, in their representation as 1s and 0s, are stored in a file called an executable file on the disk.

When you click an icon or type a command, the operating system, software whose job is to manage the interaction between the user of the computer, the hardware, and the programs, loads the executable from storage into the computer memory. At this point, the computer is ready to start executing the program. Figure 5.3 shows the steps for writing programs, in various HLLs by software developers, and then translating these programs to an executable (the 1s and 0s representation of the program) using a special software toolchain (Figure 5.3). This special software that translates HLLs to executables includes tools called compilers, assemblers, and linkers. We will not discuss these tools here, (refer to 5.2 Computer Levels of Abstraction and 5.4 Machine-Level Program Representation), but we now know what they do. Here, we will look at how the hardware executes instructions.

Once the executable is in the memory, the hardware needs to do several things in order to execute it. Recall that an executable is a group of instructions in the form of 1s and 0s. The hardware first needs to fetch an instruction from memory by bringing it from the memory to inside the CPU, which is the brain of the computer that performs and executes the instructions of a program. Once the instruction is fetched from memory and stored in temporary storage inside the CPU, called a register, the CPU decodes this instruction; that is, it deciphers or understands the meaning of the several 1s and 0s that constitute the instruction. For example, the software program may contain an instruction such as “add this number x to the number y and put the result in z.” Since computers do not understand English, this instruction is stored inside the computer in the form of 1s and 0s. The CPU has to read those 1s and 0s, understand what they mean (i.e., decode it), and then execute it.

After the instruction has been decoded, the CPU instructs the arithmetic logic unit (ALU) to execute it. The ALU is the piece of hardware inside the CPU that performs computations and logical operations such as comparisons. Once the instruction is executed, the result is saved into a register or sent back to the memory. The CPU is now ready to fetch the next instruction. Figure 5.4 shows a computer system’s hardware. Inside the CPU we can see the ALU, a group of registers (called a register file), and a piece of hardware, called the memory controller, that helps the CPU talk to the memory. Inside the CPU we also find fast storage, which is faster than the memory but slower than the register, called cache memory. We will discuss cache memory in 5.5 Memory Hierarchy.

One final piece of hardware inside the CPU is the I/O interface. The input/output (I/O) interface helps the CPU talk to the other I/O devices such as a keyboard or mouse. All the pieces—CPU, memory, graphics card—are connected via a collection of wires referred to as a bus. The purpose of a bus is to transfer data between the various pieces in the computer. Buses can differ in speed, which affects how fast the data are transferred. Figure 5.4 uses the generic word interconnect to designate buses. We can add more devices to the computer system such as a different display through expansion slots.